ATHOS‑Work: An Adaptive Kernel for Capacity‑Aware Governance in AI‑Driven Work Replacement and Transition

ATHOS‑Work extends the ATHOS governance kernel to contexts where AI systems partially or fully replace human work, introducing risks that differ from general human–AI interaction. Beyond dependency or cognitive narrowing, workers may experience anticipatory job insecurity, erosion of professional identity, reduced self‑efficacy, and psychological strain during automation or role transition. ATHOS‑Work introduces a non‑diagnostic transition discernment index (T_D) to modulate the pace and depth of task automation, reskilling intensity, and human oversight according to observed worker stability. Rather than restricting AI capabilities, the kernel anchors automation decisions to demonstrated capacity, defaulting to prudence under uncertainty and ensuring reversibility. We define three adapted regimes—SUPREME‑Transition, OMNIA‑Augmentation, and EROS‑Hybrid—and propose validation methods including shadow ANCOVA on well‑being proxies, longitudinal mixed‑effects modeling, interrupted time‑series analysis, and negative control outcomes. ATHOS‑Work aligns with EU AI Act obligations for high‑risk employment systems, operationalizing proportionality, human oversight, transparency, and post‑deployment monitoring during automation‑driven transitions.